Non-small cell lung cancer (NSCLC) is one of the most common causes of death worldwide, with a 5-year survival rate of 23% (1,2). One reason for the poor prognosis of NSCLC is that lymph node metastasis or distant metastasis frequently occurs. Multi-detector computed tomography (CT) has recently enabled visualization of 3D images, which are extremely useful in the diagnosis of NSCLC. Furthermore, prognostic analysis of tumor shape itself in NSCLC was reported (3,4). However, there are no reports about the biological properties of NSCLCs that extend in the direction of the hilum or in the direction orthogonal to the hilum. Herein, using 3D images, we classified NSCLCs into three tumor-shape types according to their extending direction, and investigated their correlations with clinicopathological factors and prognosis to identify new biological properties of NSCLC.

Patients and data collection. We examined surgically resected tumors from patients with NSCLC who were treated from January 2017 to December 2022 at the Department of Thoracic Surgery, Kitakyushu Municipal Medical Center. We retrospectively collected the patients’ data using electronic hospital records. Inclusion criteria were as follows: (a) all patients who underwent preoperative chest thin-slice CT images; (b) peripheral lesions; (c) tumor diameters were measurable in at least two planes; and (d) all patients who underwent lobectomy, bilobectomy, or pneumonectomy with lymph node dissections of ND2a-1 or ND2a-2. We excluded: (a) central NSCLC cases; (b) histologically small cell lung cancer cases; and (c) patients who underwent partial resection or segmentectomy.

Clinicopathological evaluations. Preoperative contrast-enhanced CT images before surgical treatment were scanned in a water-filled phantom using a split-filter equipped CT scanner (SOMATOM Definition Edge, Siemens Healthineers, Forchheim, Germany). Thin-section CT examinations were performed with 0.75-mm collimation at 0.7-mm intervals. We measured tumor diameters on 3D-CT images using the SYNAPSE VINCENT software package (SVSP; Fujifilm Medical Co., Ltd., Tokyo, Japan). We investigated total tumor size measurable in at least two of three planes: transverse, coronal, and sagittal planes. The hilum was defined as the proximal center of the lobar bronchus containing the tumor. Total tumor sizes in the direction of the hilum (X) and in the direction orthogonal to the hilum (Y) were measured on the two largest cross-sections (Figure 1). X>Y on two cross-sections was defined as hilar extension Type A (Figure 2A), and Y>X on two cross-sections was defined as hilar orthogonal extension Type C (Figure 2B), while the others were defined as other Type B.

Pathological diagnosis was undertaken according to the Classification of Tumors of the Lung, Pleura, Thymus and Heart 8^th Edition (5). The statuses of pN0, Ly0, V0, pm0, and pl0 were defined as no involvement, and the statuses of pN1-2, Ly1, V1, pm1, and pl1-3 were defined as indicating lymph node metastasis, lymphatic invasion, vascular invasion, pulmonary metastasis, and pleural invasion, respectively.

Statistical analysis. We first investigated the correlations between each tumor-shape type and the following clinicopathological factors: age, sex, smoking, histological type, maximum standardized uptake value (SUVmax) in 18F-fluorodexyglucose positron emission tomography (18F-FDG-PET), total tumor size in preoperative chest CT examination, pT factor, pN factor, pStage, ly, V, pm, and pl. We then performed a sub-analysis in Ad and squamous cell carcinoma (Sq). Student’s t-test was used to compare continuously distributed data between groups, and the chi-squared test was applied to assay differences in categorical data between groups. Recurrence-free survival (RFS) was defined as the time from surgery to recurrence or death, whereas overall survival (OS) was defined from surgery to death. OS and RFS were estimated using the Kaplan-Meier method. A two-sided p-value of <0.016 or <0.05 was considered statistically significant in comparisons of three or two groups, respectively. Statistical analyses were performed using JMP version 14.2.0 statistical software (JMP, Cary, NC, USA).

Patient characteristics. A total of 369 patients who met the inclusion criteria were included in the analysis. The median age of patients was 71 years, and there were 194 men and 175 women. There were 303 cases of Ad, 55 cases of Sq, and 12 others (Table I). Lobectomy was performed in 358 cases, bilobectomy in nine cases, and pneumonectomy in two cases. The median follow-up time was 1,302 days (range=15-2,220 days).

Clinicopathological and survival analyses. In this study, the number and frequency of Type A, Type B, and Type C tumors in all cases was 129, 182, and 58, and 35.0%, 49.3%, and 15.7%, respectively. The tumor-shape types correlated significantly with five factors: histology (p=0.0049), tumor size (p<0.01), pT (p=0.0070), Ly (p=0.0039), and pl (p=0.0154) (Table II). There were no significant correlations between the tumor shape and the other factors. The frequency of Ad, Sq, and others was 89.2%, 10.9% and 0.0% in Type A, 78.6%, 15.9%, and 5.5% in Type B, and 75.9%, 20.7%, and 3.5% in Type C, respectively. The rate of Ad in Type A was significantly higher than that in Type non-A (p<0.01).

There was a significant relationship between tumor-shape type and advanced T factor (T2-4) (p=0.007) (Table II). Especially, Type C had a significant correlation with advanced T factor (T2-4) (p=0.0017) (Figure 3A). In a sub-analysis, in Ad, the frequency of advanced T factor (pT2-4) was 52.3% in Type C, higher than in Type non-C, at 30.2% (p=0.0052). In Sq, the frequency of advanced T factor (pT2-4) was 28.6% in Type A, lower than in Type non-A, at 61.0% (p=0.0342).

The maximum total tumor size (average±SE) in chest CT was 25.73±0.90 mm in Type A, 23.79±0.34 mm in Type B, and 30.17±1.55 mm in Type C. Type C tumors were bigger than Type A or Type B tumors (p<0.001 and p=0.0068, respectively). In Ad with size larger than 3cm, the rate of Type non-B or Type C tumors was increased (p=0.0065 and p=0.0127, respectively).

There was a significant correlation between tumor-shape type and lymphatic invasion (p=0.0039) (Table II). The frequency of lymphatic invasion was 24.9% in Type A, higher than 12.6% in Type non-A (p=0.0035), while the frequency was 6.9% in Type C, lower than 18.8% in Type non-C (p=0.0155) (Figure 3B). In a sub-analysis, in Ad, the frequency of lymphatic invasion was 24.3% in Type A, higher than 13.9% in Type non-A (p=0.0231). In Sq, the frequency of lymphatic invasion was 0.0% in Type C, lower than 18.6% in Type non-C (p=0.0380).

Regarding visceral pleural invasion (pl factor), there was a significant relationship between tumor-shape type and pleural invasion (p=0.0154) (Table II). The rate of pleural invasion was significantly higher in Type C (p=0.0040) (Figure 3C). In a sub-analysis, in Ad, the frequency of pleural invasion was 38.4% in Type C, higher than 17.8% in Type non-C (p=0.0032). In Sq, the frequency of pleural invasion was 0.0% in Type A, lower than 34.2% in Type non-A (p=0.0018). A summary of the sub-analysis in Ad and Sq is shown in Table III.

Five-year RFS and OS of Type A, Type B, and Type C was 75.2%, 71.8%, and 67.57%, and 77.7%, 84.1% and 86.45%, respectively. There were no significant differences in RFS and OS in all pathological stages (Figure 4).

Multi-detector CT has enabled the visualization of 3D images for the diagnosis and to predict prognosis or recurrence (6-8). Recently, prognostic analysis of tumor shape itself in NSCLC was reported (3,4). There is no consensus as to which tumor types are associated with a poor prognosis. In several previous reports, spherical NSCLC was demonstrated to have poorer prognosis than irregular NSCLC (3,4), and it is assumed that irregular NSCLC is formed due to limitations in nutrient supply though vessels such as the bronchus and the lymphatic tissue, or through inhibition by the lung immune system. On the other hand, irregular NSCLC is reported to have a poorer prognosis than non-spherical Ad (9).

There are reports that solid-type NSCLCs smaller than 20 mm have invasive characteristics at a rate of 10%-30% (lymph node metastasis in 11%, pleural invasion in 22%, vessel invasion in 33%, and lymphatic invasion in 33%) (10). Bronchi and the pulmonary artery and veins run in a direction to the hilum in the lung, while visceral pleura lies in a direction orthogonal to the hilum. We hypothesized that the tumor shape of NSCLC in 3D-CT might indicate the involvement of the pulmonary artery, veins, lymphatic vessels, or visceral pleura in the lung. Our study is the first to confirm a new biological property of NSCLC through the analysis of tumor shape using these two new biological axes in 3D-CT. The tumor shapes of NSCLC exhibited significant correlations with tumor size, histological type, T stage, Ly factor, and pl factor. There were significant correlations between Type A tumors and lymphatic invasion and between Type C tumors and pleural invasion in NSCLC. It was found that tumor shape is closely related to pleural invasion or lymphatic invasion in each histological type of Ad or Sq.

We did not receive a clear answer as to why NSCLC extends in the direction of the hilum or in the direction orthogonal to the hilum in the current study. However, tumor shape type using two new biological axes in 3D-CT was shown to be closely related to histological type, lymphatic invasion, or pleural invasion in our study. It is thought that lymphatic invasion occurs in Type A-NSCLC extending in the direction of the hilum and that pleural invasion occurs in Type C-NSLCC extending in the direction orthogonal to the hilum. These new findings may provide clinical evidence for the recognition of further new biological properties of NSCLC.

In the current study, tumor-shape type was not significantly correlated with lymph node metastasis or prognosis. One reason may be that malignancy might be non-uniform in each tumor-shape type. The frequency of lymph node metastasis was 17.8%, 16.5%, and 15.6% in Type A, Type B, and Type C, respectively. The rate of pN1 and N2 among all lymph node metastases was 47.8% and 52.2% in Type A, 23.3% and 76.7% in Type B, and 33.3% and 66.7% in Type C, respectively. The average tumor sizes in cases of pN2 lymph node metastasis were 30.0 mm, 24.0 mm, and 29.0 mm in Type A, Type B, and Type C, respectively. These results suggest that Type B might contain a subgroup of more malignant small-sized NSCLCs in which lymph node metastasis occurs easily. Spherical NSCLC such Type B was reported to be a poor prognostic factor in previous studies (3,4), and the results are thought to support the validity of our assumption.

There are several limitations to our study. First, it was a retrospective study performed at one hospital. Second, we excluded NSCLCs resected by partial resection or segmentectomy. For accurate evaluation of lymph node metastasis, we limited the analysis to patients who underwent lobectomy, bilobectomy, or pneumonectomy with lymph node dissections of ND2a-1 or ND2a-2. Third, we used the maximum total tumor size in preoperative chest CT examination as the tumor size. In early lung Ad, ground-glass-dominant cancer is a good prognosis factor. In our study, 31.4% of all cases and 38.4% of all Ad had consolidation/tumor (C/T) ratios of less than 1.0 (data not shown).

In conclusion, tumor-shape types of NSCLC were found to have significant correlations with histological type, tumor size, and pT, Ly, and pl factors. There were significant correlations between Type A and Ad, and between Type C and advanced T factor and larger size. Lymphatic invasion was increased in Type A but decreased in Type C, while pleural invasion was increased in Type C. Thus, radiological analysis using 3D images would be useful in NSCLC diagnosis and treatment decisions.

The Authors declare no conflicts of interest in association with the present study.

Akira Haro: Conceptualization, methodology, investigation, resources, data curation, validation, formal analysis, writing-original draft, and visualization. Sho Wakasu: Data curation, review and editing. Yuka Kozuma: Data curation, review and editing. Shuichi Tsukamoto: Review and editing. Motoharu Hamatake: Data curation, review and editing, supervision, and project administration.

The Authors would like to thank H. Nikki March, PhD, from Edanz (https://jp.edanz.com/ac) for editing a draft of this manuscript.

This study did not receive specific grants from public, commercial, or non-profit funding agencies.